Skip to content

armgit / viz3d

Public Imported from GitHub
0

An AI agent skill that turns any subject into a 3D film or app

Brief: interpolate fixed-step motion, no overlapping or coplanar surfaces

Nguyen Xuan Ana7f1a620b16d
diff --git a/plugins/visualize-in-3d/skills/visualize-in-3d/references/brief.md b/plugins/visualize-in-3d/skills/visualize-in-3d/references/brief.md
index 79ac6f2..17d3c00 100644
--- a/plugins/visualize-in-3d/skills/visualize-in-3d/references/brief.md
+++ b/plugins/visualize-in-3d/skills/visualize-in-3d/references/brief.md
@@ -78,7 +78,12 @@ Work only in {{FOLDER}}/app. Stack: Vite + React + React Three Fiber (+ drei, @r
Playwright as a dev dependency for checks and the film, all pinned to exact versions, serve on port
{{PORT}}. A plain data model (sites, units, tasks, events) drives everything; the 3D view and the dashboard both read
it, and a small seeded simulation advances it in real time so things move, statuses change and numbers tick. Load the
-GLBs with instancing for repeats, drive every moving part and clip from the simulation. Clicking an object selects it
+GLBs with instancing for repeats, drive every moving part and clip from the simulation. A fixed-step simulation
+keeps the previous and current pose of everything that moves and renders the pose interpolated by the leftover
+time, so motion and the follow camera change every frame (no stutter). No two surfaces share a plane and no two
+modules overlap: lay repeats end to end, build curves (track, viaduct deck) as continuous swept geometry, lift
+stacked floors a few centimetres apart, and set the camera near plane to about 0.5 m. Layered transparent surfaces
+(glass over glass) read as a smear; keep one glass layer and make the rest opaque. Clicking an object selects it
and opens its card; clicking a list row flies the camera to it (and follows it if it moves); switching sites changes
the world. Soft shadows, ambient occlusion, gentle bloom on lights, smooth camera moves. 60 fps at 1920x1080 on a
mid-range GPU such as an RTX 3060.
diff --git a/plugins/visualize-in-3d/skills/visualize-in-3d/references/film.md b/plugins/visualize-in-3d/skills/visualize-in-3d/references/film.md
index 97a97ac..faf816e 100644
--- a/plugins/visualize-in-3d/skills/visualize-in-3d/references/film.md
+++ b/plugins/visualize-in-3d/skills/visualize-in-3d/references/film.md
@@ -90,5 +90,6 @@ leans on them most. Reading map: Pacing and text, Camera, Figures and places, Ma
- Look for: captions cut early, figures under captions, clipped labels, empty map areas, soft shots, floating or
sunken figures, doors that open onto walls, vehicles backing along their route, black frames at cuts. In a live app
film also look for: the cursor hidden or off target, UI over the action, selection rings wider than the object, a
- followed unit that never reaches its event, and an event too short to see.
+ followed unit that never reaches its event, an event too short to see, motion that steps every few frames
+ (compare two consecutive frames), and surfaces that flicker or overlap (crop and zoom full-size frames).
- Re-render only the time ranges you changed when the renderer supports it, then re-encode and re-check those ranges.